Article Analysis of Melanin Structure and Biochemical

https://doi.org/10.1167/tvst.8.3.33

Article

Analysis of Melanin Structure and Biochemical Compositionin Conjunctival Melanocytic Lesions Using Pump–ProbeMicroscopy

Francisco E. Robles1,2, Sanghamitra Deb1,3, Lejla Vajzovic4, Gargi K. Vora4, PrithviMruthyunjaya4,5, and Warren S. Warren1,6,7

1 Department of Chemistry, Duke University, Durham, NC, USA2 Wallace H. Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, Atlanta, GA, USA3 Beckman Institute for Advanced Science and Technology, University of Illinois at Urbana-Champaign, Urbana, IL, USA4 Department of Ophthalmology, Duke University Medical Center, Durham, NC, USA5 Department of Ophthalmology, Stanford University Medical Center, Palo Alto, CA, USA6 Department of Radiology, Duke University Medical Center, Durham, NC, USA7 Duke University, Departments of Physics, Durham, NC, USA

Correspondence: Warren S. Warren,Duke University, Chemistry, Durham,NC 27708, USA. e-mail: [email protected]

Received: 11 October 2018Accepted: 15 April 2019Published: 4 June 2019

Keywords: conjunctival melanoma;primary acquired melanosis; be-nign conjunctival nevus; pump–probe microscopy; multiphotonmicroscopy

Citation: Robles FE, Deb S, VajzovicL, Vora GK, Mruthyunjaya P, WarrenWS. Analysis of melanin structureand biochemical composition inconjunctival melanocytic lesions us-ing pump–probe microscopy. TransVis Sci Tech. 2019;8(3):33, https://doi.org/10.1167/tvst.8.3.33Copyright 2019 The Authors

Purpose: We analyze melanin structure and biochemical composition in conjunctivalmelanocytic lesions using pump-probe microscopy to assess the potential for thismethod to assist in melanoma diagnosis.

Methods: Pump–probe microscopy interrogates transient excited-state photodynamicproperties of absorbing molecules, which yields highly specific molecular informationwith subcellular spatial resolution. This method is applied to analyze melanin in 39unstained, thin biopsy specimens of melanocytic conjunctival lesions. Quantitativefeatures of the biochemical composition and structure of melanin in histopathologicspecimens are assessed using a geometric representation of principal componentanalysis (PCA) and principles of mathematical morphology. Diagnostic power isdetermined using a feature selection algorithm combined with cross validation.

Results: Conjunctival melanomas show higher biochemical heterogeneity anddifferent overall biochemical composition than primary acquired melanosis of theconjunctiva (PAM) without severe atypia. The molecular signatures of PAMs withsevere atypia more closely resemble melanomas than other types of PAMs. Pigmentorganization in the tissue becomes more disorganized as diagnosis of the lesionsworsen, but nevi are more inconsistent biochemically and structurally than otherlesions. Relatively high sensitivity (SE) and specificity (SP) is achieved fordifferentiating between various melanocytic lesions, particularly PAMs without severeatypia and melanomas (SE ¼ 89%; SP ¼ 87%).

Conclusions: Pump–probe microscopy is a powerful tool that can identify quantitative,phenotypic differences between various types of conjunctival melanocytic lesions.

Translational Relevance: This study further validates the use of pump–probemicroscopy as a potential diagnostic aid for histopathologic evaluation of conjunctivalmelanocytic lesions.

Introduction

Melanocytic proliferations in the conjunctiva

account for more than half of all neoplasms in that

anatomic region, and fall under three broad catego-

ries: melanocytic nevus, conjunctival melanosis

(which includes complexion-associated melanosis

and primary acquired melanosis [PAM], also called

conjunctival melanocytic intraepithelial neoplasia [c-

MIN]1), and invasive conjunctival melanoma.2 Inva-

sive melanomas account for 2% of all ocular

1 TVST j 2019 j Vol. 8 j No. 3 j Article 33

This work is licensed under a Creative Commons Attribution 4.0 International License.Downloaded from tvst.arvojournals.org on 06/04/2019

https://creativecommons.org/licenses/by/4.0/

malignancies3 and, like cutaneous melanomas,4 theirincidence rate is rapidly growing; for white men in theUnited States, for example, the indicine rate increasedby 295% from 1973 to 1999.5 Conjunctival melano-mas typically develop de novo (12%–47% of the time6;these also are considered the most aggressive5,6), orevolve from a subtype of conjunctival melanosis(10%–74%).3,6–8 Conjunctival nevi are benign prolif-erations that are classified similarly to cutaneous neviand more rarely evolve to malignant melanomas (1%–26%).6,8

Accurate clinical diagnosis of a conjunctivalmelanocytic lesion can be challenging for a numberof reasons. For example, conjunctival melanocyticproliferations contain histologic patterns that differsubstantially from their cutaneous counterparts, butthe pathology-based terminology used to definecutaneous melanocytic proliferations often are ap-plied by nonocular pathologists to ocular specimens,which are not always applicable.1,9 Also, PAM isfurther subcategorized histopathologically by severityof atypia (no atypia, mild, moderate, or severe); whilethis terminology maybe useful for monitoring theprogression of these lesions,1 the gradation is highlyvulnerable to the subjective nature of histopathologicinterpretation. Plus, there often exists an overlap inappearance of PAM and relatively flat or diffuseconfigurations of conjunctival melanoma. In fact,some experts argue that PAMs with severe atypiashould be considered melanomas in situ7,9,10 – thevast majority of PAMs that evolve to invasivemelanomas fall under this subcategory7,9 and shouldbe treated as such. Moreover, the most recent updateof the American Joint Committee on Cancer clinicalclassification describes the T(is) stage as conjunctivalmelanoma in situ, or melanoma confined to theepithelium. This condition includes primary acquiredmelanosis with atypia in more than 75% of normalepithelial thickness.11 These considerations compli-cate reliable and reproducible clinical and histopath-ologic evaluation.

Noninvasive imaging technologies have the poten-tial to aid in the diagnosis of these melanocyticlesions, but to date such imaging interventions havebeen limited. For example, confocal microscopy,12

optical coherence tomography,13 and ultrasound14

have been used to provide three-dimensional (3D),tomographic images of the structure of conjunctivallesions; but the source of contrast for these methods,optical or acoustic scattering, fails to provide directdisease-specific information that can help assess thepathology. Thus, their use has not been widely

adopted (but continue to be studied). To alleviatethese shortcomings, we have developed pump–probemicroscopy, a novel label-free molecular imagingmethod that provides unique insight into the bio-chemical composition of melanin, which serves as anindicator of melanocyte activity that can be used as abiomarker of disease.15–17

In short, pump–probe microscopy applies a two-color pulsed laser system to interrogate the ultrafasttransient excited and ground state photodynamicproperties of pigments. Implementation of thisapproach is related to laser scanning microscopy,but yields highly specific biochemical informationwithout exogenous agents. We have shown thatpump–probe microscopy is sensitive not just to thetype of melanin (eu- and pheo-melanin), but alsooxidation state, aggregate size, and metal content.18

Using thin, unstained biopsy specimens, we leveragedthis biochemical sensitivity to quantitatively differen-tiate melanomas from melanocytic nevi and toprovide novel insight into the metastatic potential ofinvasive melanomas in the skin15,16,19 and femalegenital tract.17 Using a small sample set (N ¼ 8), wealso showed that pump–probe microscopy canidentify average differences in pigmentation patternsamong conjunctival nevi, PAMs, and melanomas, andthat pigment in the conjunctiva is not identical to thatfound in cutaneous lesions.20

In this work, we significantly increased our sampleset (N ¼ 39) to better evaluate the diagnostic use ofpump–probe microscopy for melanocytic prolifera-tions in the conjunctiva. We applied a geometricrepresentation of principal component analysis (PCA)and used principles of mathematical morphology toquantify the molecular signatures alongside thespatial structure of the pigment. A robust statisticalanalysis was applied to assess the ability of pump–probe microscopy to classify conjunctival melanocyticlesions. Results demonstrated that this novel molec-ular imaging method is a promising tool fordifferentiating various types of conjunctival lesions,particularly PAMs without severe atypia and mela-nomas. Data also indicated that PAMs with severeatypia share a close pigment phenotypic relationshipwith melanomas, providing some additional supportto the notion that the former lesions may be bettercategorized as melanomas in-situ.

Methods

We analyzed 39 previously excised conjunctivalmelanocytic lesions (Table 1). From each fixed tissue


Robles et al.

Downloaded from tvst.arvojournals.org on 06/04/2019

specimen, two adjacent, thin (~10 lm) sections werecut, where one was left unstained while the other wasstained with hematoxylin and eosin (HE). Anophthalmic pathology expert used the HE-stainedsections to classify regions of interest; then, theseregions were identified in the adjacent, unstainedsections and imaged with pump–probe microscopy. Atotal of 471 pump–probe images were acquired. Thesestudies were conducted under a protocol approved bythe Duke University School of Medicine institutionalreview board.

The pump–probe system has been described indetail previously.15,21,22 Briefly, we used a custom-built laser scanning microscope, equipped with a dual-output (i.e., two color) femtosecond laser system. Thetwo outputs were derived from a Coherent Miraoptical parametric oscillator (OPO; Coherent, Inc.,Santa Clara, CA) and a 80 MHz ChameleonTi:Sapphire oscillator (Coherent, Inc.), tuned to 730and 810 nm, respectively serving as the pump andprobe pulses. The laser pulses used less than 0.65 mWtotal power to avoid any effects that may alter themelanin pigment chemistry (e.g., photobleaching).18

Molecular signatures were obtained by firstmodulating the 730-nm pump beam at 2 MHz, whileleaving the 810-nm probe beam unmodulated. Bothbeams were spatially overlapped and then sent

collinearly to the microscope. When the beams werefocused onto the sample, nonlinear optical interac-tions between the pump and probe (which interrogateelectronic excited and ground states of the sample)transferred the 2-MHz modulation to the probe. Thissignal then was detected with a photodiode and alock-in amplifier. Finally, the ultrafast photodynamicproperties were obtained by varying the time-delaybetween the pump and probe pulses. A total of 34time-delays, ranging from approximately 0 to 4picoseconds (ps), were used to determine the photo-dynamic properties. The data acquired were 3D, withtwo dimensions being the x and y coordinates of thethin sample, and the third being the pump–probetime-delays, which yielded the ultrafast transientdynamics (i.e., molecular signatures; Fig. 1). Eachindividual pump–probe image had a field of view of3603 360 lm.

To quantify the molecular signatures, we used ageometric representation of principal component (PC)analysis.16,23,24 This procedure (Fig. 1) allowed us toreduce the dimensionality of the data set (from 34time-delays to only a few PCs) for quantitativeanalysis and enabled a visual representation of thesignals. Here, we leveraged the fact that .97% of thesignal variance could be described by the top 3 PCs.This could be represented in a spherical coordinatesystem, where the angles (h and u) describe thetransient dynamics, while the radius (R) representsthe signal strength (i.e., relative concentration). Forimage display, we used a hue saturation value colormap scheme, where the hue was set based on h (Fig.1c, inset), the value is set by R, and the saturation isset to 1.

Results

Pump–Probe Microscopy Images ofMelanocytic Proliferations in the Conjunctiva

Pump–probe signals arising from melanin (exclud-ing surgical ink and hemoglobin) from the entiresample set were first processed with PCA to derive thetop PCs. As the inset of Figure 1b shows, the top 3PCs correspond to .97% of the variance, which thenwere used to construct a visual representation of thepump–probe signals (Fig. 1c) based on the angles ofthe PCs spherical coordinate system, as describedabove. As Figure 1c shows, the two-dimensional (2D)histogram of the entire data set reveals two majorcomponents (also known as endmembers),16,23 withall other signals comprising a linear combination of

Table 1. Sample Set Characteristics

No. ofImagesa

Sectionswith Tissue

TypePatient

Diagnosis

Nevi 145 16 16PAM w/o atypiaand w/ mild/modatypia

85 10 9

PAM w/ severeatypia

47 4 3

Melanoma 194 10 11Total 471 40 39

Using PAM with severe atypia as an example, the sampleset contained three specimens from patients diagnosedwith PAM with severe atypia, but four total sections in thestudy contained indications of PAM with severe atypia. Theextra section with this tissue type came from a patient withmelanoma whose tissue has regions more consistent withPAM with severe atypia. A total of 47 images were takenfrom regions with PAM with severe atypia. w/o, without; w/,with; mod, moderate.

a Each image region is a 3603 360 lm2 area.


Robles et al.


these two components. One endmember is located atan azimuth angle of h ¼ 0 rad (i.e., a positivecontribution from PC1), which signifies a loss ofnonlinear interaction (e.g., excited state absorption).These types of signals have been observed in melaninswith large aggregates and those void of metals.18 Theother endmember lies at approximately h ¼ 2.25 rad(negative contribution from PC1, and a positivecontribution from PC2) and indicates a gain in theprobe from interactions, like ground-state bleaching.These signals have been observed in melanins withhigh levels of metal content, small aggregates, andsynthetic pheomelanin.18,25 Thus, the pump–probesignals, and hence the mapping of the signals onto thisprincipal component space, gives an indication ofmelanin biochemical composition (albeit not unique-ly). A more detailed analysis on how the pump–probesignals differ based on chemical composition has beenreported previously.18,25 The mapping of each spatialpixel’s molecular signature on the h-axis also definesthe color assignment in the pump–probe images – thecolor map is illustrated in Figure 1c (horizontal colormap).

Figures 2 to 6 illustrate selected pump–probeimages of unstained tissue sections, along with their2D histogram of the PCs angular distribution and theadjacent H&E stained sections. Figure 2 shows abenign nevus with representative characteristics ofthese lesions in our sample set. However, as wediscuss below, these lesions show a wide spectrum offeatures, which is a hallmark of nevi.9,11 This lesion ishighly pigmented with a moderate level of biochem-

ical heterogeneity, as represented by the differentcolors in the pump–probe image and the spread of thesignals in the 2D histogram. The pigment is primarilyin the substantia propria, with a high degree ofstructural heterogeneity, meaning that the pigment isnot well organized or confined to a particular area.Figure 3 illustrated a different type of benign nevus, ablue nevus. This unique lesion shows a biochemicalcomposition unlike any other investigated to date,with a strongly bimodal distribution. One cluster,which appears to correspond to spindle melanocytesor melanin pigment around the fibrotic stroma, showsa distinct pink hue from pump–probe signals thatclosely resemble the first pc (signifying a lossnonlinear interaction indicative of large melaninaggregates and melanins void of metals). The othercluster, in cyan, has the opposite pump–proberesponse – mostly a negative contribution from thefirst PC, signifying a gain in the probe pulse (resultingfrom melanin with high levels of metal content, smallaggregate size, or pheomelanin). The morphologysuggests that this response is from melanin in thecytoplasm of plump, epithelioid melanocytes, whichare characteristic of this type of lesion.10,26

Next, we showed two PAMs. The first (Fig. 4) is aPAM with mild-to-moderate atypia, with the tissuesection containing two separate segments (from thesame patient) that exhibit slightly different biochem-ical characteristics. Note that while the color of eachsegment appears drastically different, in fact themolecular signatures are only slightly different asseen by the small separation of the two clusters (one

Figure 1. (a) Illustration of pump–probe microscopy data. Data acquired to generate pump–probe images are 3D, with two spatialdimensions (x and y) and the third consisting of the transient photodynamics. (b) Transient signals from melanin are processed withprinciple component analysis, where the top three principal components (pc’s) capture over 97% of the data variance. (c) 2D histogramof melanin transient signals. The top three pc’s span a 3D space, whose angles (in spherical coordinates) can be used to graphicallyrepresent the biochemical composition, as shown in the inset.


Robles et al.


from each segment) on the same 2D histogram. Inboth pieces, the pigment is found diffusely throughthe epithelium with relatively elevated concentrationsin the basal layer. Overall, the lesion shows a higherdegree of organized pigment structure compared tothe nevi, meaning that the pigment is confined to aparticular region and it resembles the structure ofnormal tissue. The faint dark blue circles in the uppersegment are from hemoglobin, which have h ¼ 0. Thesecond PAM (Fig. 5) contains severe atypia, andexhibits a more disorganized and widespread pigmentstructure, with a few scattered cells possessingdifferent pump–probe (and, thus, molecular) signa-tures compared to the majority of the surroundingpigment. In this image, the cyan color corresponds tosurgical ink. Note that the elevation angle, /, isdifferent for the ink than the melanin seen in the bluenevus even though both are mapped to a cyan hue.

Finally, Figure 6 shows an invasive melanoma.Here, the pigment is highly heterogeneous, structur-ally and biochemically. The pigment is in theepithelium and substantia propria, and has a finertexture compared to the coarser pigment structuresfound in the nevi and PAMs without severe atypia.

Pigment in the cytoplasm also reveals many cells withlarge nucleus and high nucleus-to-cytoplasm ratio.

Image Quantification and MultivariateAnalysis

To quantify different features of the conjunctivallesions, we used parameters derived from the melaninbiochemical composition and its structure. Averageparameters of the biochemical composition of thelesions (i.e., that do not take structure into account)are readily assessed using the 2D PC histograms,where we extract the mean, standard deviation, andentropy of the distribution. Then, to account for theoverall structure and structure of each endmember,we use mathematical morphology.16,19 In this process,three gray scale images were produced. The firstconsisted of the total melanin pigment obtained fromthe magnitude of the signal (R in Fig. 1c), whichsuppresses the detailed biochemical information. Theother two gray-scaled imaged are weighted by therelative concentration of the two endmembers. Forone, the R image is multiplied by 1 for pixels with h¼2.5 and 0 for pixels with h¼ 0 rad (denoted hþ), andthe other uses the opposite scale (denoted h�). These

Figure 2. Pump–probe microcopy image of unstained benign nevus biopsy sample, along with its 2D histogram of the pc’s angulardistribution and adjacent, HE-stained microphotograph.


Robles et al.


Figure 3. Pump–probe microcopy image of unstained benign blue nevus biopsy sample, along with its 2D histogram of the pc’sangular distribution and adjacent, HE-stained microphotograph.

Figure 4. Pump–probe microcopy image of PAM with mild-to-moderate atypia, along with its 2D histogram of the pc’s angulardistribution and adjacent, HE-stained microphotograph.


Robles et al.


Figure 5. Pump–probe microcopy image of PAM with severe atypia, along with its 2D histogram of the pc’s angular distribution andadjacent, HE-stained microphotograph.

Figure 6. Pump–probe microcopy image of invasive conjunctival melanoma, along with its 2D histogram of the pc’s angulardistribution and adjacent, HE-stained microphotograph.


Robles et al.


two images capture the structural and biochemicalinformation of the melanin. Finally, we applied a 2Dmathematical autocorrelation transformation andextracted second-order geometric information (i.e.,morphologic covariance) to quantify the structure.Details of this method have been reported previous-ly.19 Overall, 30 features are extracted from eachpump–probe image.

Finally, we applied a multivariate analysis consist-ing of a feature selection algorithm to identify thefeatures with the most diagnostic value, followed bycross-validation. Details for this method have beendescribed previously by Robles et al.16 In short,forward sequential feature selection was used in awrapper fashion with data drawn from individual

images to capture the most amount of variationacross the data. Once features were selected, weassessed the diagnostic power (sensitivity [SE] andspecificity [SP]) by applying the leave-one-out crossvalidation (LOOCV) method, with all images fromone patient treated as one ‘sample’ so that they arewithheld from training to ensure that the training setwas independent of the test set. While LOOCV doeshave its limitations, including possible bias due to apriori knowledge of the diagnosis, it is a well-acceptedmethod to obtain some measure of predictive powerwhen the sample size is small.

Figure 7 shows the most important features thatdistinguish nevi, PAMs without severe atypia, PAMswith severe atypia, and melanomas – the featurescorrespond to a combination of average biochemicalcomposition (left column) as well as the pigmentstructure (right column). The data present severalinteresting trends, starting with the apparent uniquebiochemical composition of PAMs without severeatypia. Here, they show a higher h-value, signifying amore negative pump–probe response compared tomelanomas. This also is reflected by a higher value ofthe second PC. Both results are consistent with ourprevious finding in the smaller pilot study.20 Thelower h entropy for PAMs compared to melanomassuggests a lower biochemical heterogeneity, which isconsistent with the observations provided above.Interestingly, PAMs with severe atypia exhibit anaverage pigment composition that more closelyresembles melanomas instead of PAMs without severeatypia. Nevi also show a very wide and heterogeneousbiochemical composition, consistent with our previ-ous observations.

The structural analysis also revealed an interestingprogression from PAMs without severe atypia tomelanoma. Specifically, PAMs without severe atypiahave the highest level of anisotropy in the R and himages, which describes its well-bounded pigment inthe epithelium. As the atypia worsens and thenprogresses to melanoma, the pigment spreads andresults in lower levels of anisotropy. Similar trendshave been observed for cutaneous and vulvarpigmented lesions.17 The nevi, however, did notfollow this trend, and again showed much highervariability in their pigment structure. The laststructural parameter, R signal-to-noise ratio (SNR),is a measure of the overall level of pigmentation(ignoring biochemical composition), and showed thatmelanomas and nevi express the most pigment.

Table 2 presents the LOOCV results. PAMswithout severe atypia can be differentiated from

Figure 7. (a–f) Box plots summarizing selected features used todifferentiate between various conjunctival melanocytic lesions.Features on the left column (a–c) only take the biochemicalcomposition into account, while features on the right column (d–f)are derived from the structure and biochemical composition ofmelanin.


Robles et al.


melanomas with high SE (89%) and SP (87%).However, it remains difficult to separate the PAMswith severe atypia from either melanoma (SE¼ 87%,SP ¼ 57%) and from PAMs without severe atypia(SE¼ 55%, SP¼ 85%). This problem likely is due tothe low number of lesions available with PAMs withsevere atypia, but also is compounded by the subtlecontinuum that differentiates these lesions. A largersample set could reveal a more robust set of featuresthat help assess quantitative differences betweenthese subgroups. Separation of the nevi from PAMswithout severe atypia is fairly robust, but degradesslightly for PAMs with severe atypia and melano-mas.

Discussion

We have shown that pump–probe microscopy is apowerful tool with the potential to aid in the diagnosisof conjunctival melanocytic lesions. Results showedthat melanomas have higher biochemical heterogene-ity and different average biochemical compositionthan PAMs without severe atypia. The molecularsignatures of PAMs with severe atypia, on the otherhand, more closely resemble melanomas than othertypes of PAMs, in support of the notion that theselesions may be better categorized as melanomas in-situ. Structurally, the pigment becomes more disor-ganized as diagnosis of the lesions worsened. Dataalso showed that nevi have a more inconsistentbiochemical and structural composition than otherlesions. Using six (of 30) parameters that quantify thestructure along with the biochemical composition ofthe melanin pigment, we demonstrated a relativelyhigh sensitivity and specificity for differentiatingbetween various melanocytic lesions, particularlyPAMs without severe atypia and melanomas.

This study focused on excised, thin tissue biopsyspecimens, and as such we have shown that pump–probe microscopy can serve as a quantitative tool to

achieve a more accurate diagnosis, one rooted on anendogenous molecular phenotype (i.e., melanin ex-pression) that indicates malignancy. While this studyused a moderate sample size for conjunctival mela-nocytic lesions, a larger and, more importantly,prospective study will be needed to more definitivelydetermine the diagnostic use of pump–probe micros-copy.

It also is important to note that while theprocessing and/or fixing protocols of these samplesmay differ, the pump–probe signals from melanin areremarkably stable. Data (not shown) indicated thatthe pump–probe signals (from human brown hair)remain invariant to different duration of fixation andmaterials. These results are further supported by thefact that all of the (hundreds of) samples we tested(from skin, conjunctiva, and so forth), which havesurely not been processed in the exact same way,possess the same general set of characteristics. This isnot too surprising given that the pump–probe signalscome from melanin, not associated with proteins thatare crosslinked in the fixation process; thus, thefixation has much less effect on the melanin pump–probe signals. We also showed that the pump�probesignature of eumelanin from the ink sac of a Jurassiccephalopod is identical to that of its modern-daycounterpart,27 further validating the stability ofmelanin, and pump–probe signals, for analyzingarchived pigmented tissue.

Finally, in vivo imaging with pump–probe micros-copy is possible,28 albeit with more limited access tothe biochemical composition,29 and in the future itmay provide similar information without the need forexcisional biopsy. Our study not only will serve as aroadmap to future endeavors for in vivo imaging, butit also stands alone by establishing pump–probemicroscopy as a potential diagnostic aid for histo-pathologic evaluation of conjunctival melanocyticlesions.

Table 2. Sensitivity and Specificity as Determined by LOOCV

SE, % SP, % No. Lesions

PAM w/o atypia and w/ mild/mod atypia vs. melanoma 89 87 20PAM w/ severe atypia vs. melanoma 87 57 15PAM w/o atypia and w/ mild/mod atypia vs. PAM w/ severe atypia 55 85 12Nevi vs. melanoma 72 64 27Nevi vs. PAM w/o atypia and w/ mild/mod atypia 86 83 25Nevi vs. PAM w/ severe atypia 70 76 20


Robles et al.


Acknowledgments

Supported by the National Institutes of Health(NIH; Bethesda, MD) Grant NIH R01-CA166555(WSW); Burroughs Welcome Fund 1012639 (FER);NSF CBET 1752011 (FER); the Duke Center for Invivo Microscopy NIH P41EB015897, Research toPrevent Blindness (Duke Eye Center Small Grant;LV), and Duke University.

Disclosure: F.E. Robles, None; S. Deb, None; L.Vajzovic, None; G.K. Vora, None; P. Mruthyunjaya,None; W.S. Warren, None

References

1. Damato B, Coupland SE. Conjunctival melano-ma and melanosis: a reappraisal of terminology,classification and staging. Clin Exp Ophthalmol.2008;36:786–795.

2. Zembowicz A, Mandal RV, Choopong P. Mela-nocytic lesions of the conjunctiva. Arch PatholLab Med. 2010;134;1785–1792.

3. McCartney ACE. Pathology of ocular melano-mas. Br Med Bull. 1995;51:678–693.

4. Glusac EJ. The melanoma ‘epidemic’, a derma-topathologist’s perspective. J Cutan Pathol. 2011;38;264–267.

5. Yu GP, Hu DN, McCormick S, Finger PT.Conjunctival melanoma: is it increasing in theUnited States? Am J Ophthalmol. 2003;135;800–806.

6. Shields CL, Markowitz JS, Belinsky I, et al.Conjunctival Melanoma: Outcomes Based OnTumor Origin In 382 Consecutive Cases. Oph-thalmology. 2011;118;389–395.e2.

7. Shields JA, Shields CL, Mashayekhi A, et al.Primary acquired melanosis of the conjunctiva:experience with 311 eyes. Trans Am OphthalmolSoc. 2007;105:61–71, discussion 71–72.

8. Vora GK, Demirci H, Marr B, Mruthyunjaya P.Advances in the management of conjunctivalmelanoma. Surv Ophthalmol.62,26–42 (2017).

9. Zembowicz A, Mandal RV, Choopong P. Mela-nocytic lesions of the conjunctiva. Arch PatholLab Med. 2010;134:1785–1792.

10. Mudhar HS. Melanocytic lesions of the conjunc-tiva. Diagn Histopathol. 2015;21:1–10.

11. Amin MB, Edge SB, Greene FL, et al. Malignantmelanoma of the conjunctiva. In: AJCC Cancer

Staging Manual, 8th ed. New York, NY: Spring-er; 2017.

12. Messmer EM, Mackert MJ, Zapp DM, KampikA. In vivo confocal microscopy of pigmentedconjunctival tumors. Graefe’s Arch Clin ExpOphthalmol. 2006;244:1437–1445.

13. Nanji AA, Sayyad FE, Galor A, Dubovy S, KarpCL. High-resolution optical coherence tomogra-phy as an adjunctive tool in the diagnosis ofcorneal and conjunctival pathology. Ocul Surf.2015;13:226–235.

14. Ho VH, Prager TC, Diwan H, Prieto V, EsmaeliB. Ultrasound biomicroscopy for estimation oftumor thickness for conjunctival melanoma. JClin Ultrasound. 2007;35:533–537.

15. Matthews TE, Piletic IR, Selim MA, SimpsonMJ, Warren WS. Pump-probe imaging differen-tiates melanoma from melanocytic nevi. SciTransl Med. 2011;23:7tra15.

16. Robles FE, Deb S, Wilson JW, et al. Pump-probeimaging of pigmented cutaneous melanomaprimary lesions gives insight into metastaticpotential. Biomed Opt Exp. 2015;6:3631–3645.

17. Robles FE, Deb S, Fischer MC, Warren WS,Selim MA. Label-free imaging of female genitaltract melanocytic lesions with pump-probe mi-croscopy. J Low Genit Tract Dis. 2017;21:137–144.

18. Simpson MJ, Wilson JW, Robles FE, et al. Near-infrared excited state dynamics of melanins: theeffects of iron content, photo-damage, chemicaloxidation, and aggregate size. J Phys Chem A.2014;118:993–1003.

19. Robles FE, Wilson JW, Warren WS. Quantifyingmelanin spatial distribution using pump-probemicroscopy and a 2-D morphological autocorre-lation transformation for melanoma diagnosis. JBiomed Opt. 2013;18:120502.

20. Wilson JW, Vajzovic L, Robles FE, CummingsTJ, Mruthyunjaya P, Warren WS. Imagingmicroscopic pigment chemistry in conjunctivalmelanocytic lesions using pump-probe lasermicroscopy. Invest Ophthalmol Vis Sci. 2013;54:6867–6876.

21. Fu D, Ye T, Matthews TE, Yurtsever G, WarrenWS. Two-color, two-photon, and excited-stateabsorption microscopy. J Biomed Opt. 2007;12:054004.

22. Fischer MC, Wilson JW, Robles FE, Warren WS.Invited review article: pump-probe microscopy.Rev Sci Instrum. 2016;87:031101.

23. Thompson A, Robles FE, Wilson JW, Deb S,Calderbank R, Warren WS. Dual-wavelength


Robles et al.


pump-probe microscopy analysis of melanincomposition. Sci Rep. 2016;6:1–10.

24. Robles FE, Wilson JW, Fischer MC, Warren WS.Phasor analysis for nonlinear pump-probe mi-croscopy. Opt Exp. 2012;20:17082.

25. Ju KY, Fischer MC, Warren WS. Understandingthe role of aggregation in the broad absorptionbands of eumelanin. ACS Nano. 2018;12:12050–12061.

26. Jay B. Naevi and melanomata of the conjunctiva.Br J Ophthalmol. 1965;49:169–204.

27. Simpson MJ, Glass KE, Wilson JW, Wilby PR,Simon JD, Warren WS. Pump–probe microscopicimaging of Jurassic-aged eumelanin. J Phys ChemLett. 2013;4:1924–1927.

28. Matthews TE, Wilson JW, Degan S, et al. In vivoand ex vivo epi-mode pump-probe imaging ofmelanin and microvasculature. Biomed Opt Exp.2011;2:1576–1583.

29. Wilson JW, Degan S, Gainey CS, et al. Compar-ing in vivo pump–probe and multiphoton fluo-rescence microscopy of melanoma and pigmentedlesions. J Biomed Opt. 2015;20:051012.


Robles et al.


Documents

Article Analysis of Melanin Structure and Biochemical